[EM] Fwd[ To Martin on bad consequences ]

Martin Harper mcnh2 at cam.ac.uk
Wed May 2 00:43:10 PDT 2001


MIKE OSSIPOFF wrote:

> Martin-- Was this letter went only to me, or is it an EM posting?
> If it's sent only to me, then fine. If it's an EM posting, then why
> is a separate copy sent to me by direct e-mail?

My bad - my mailreader seems to deal with replys to your posts oddly, 
and sometimes I forget to correct that...  :-( I'm forwarding both 
original and reply to the list.

--

Consequence One: People discover that the optimum strategy is to always 
give a full ranking, even where you have no or very little preference, 
and decide it's all too much effort. They don't turn up. Turnout drops. 
The wrong candidate is elected.

Mike Comments: Sorry, but Consequence One isn't valid. Perhaps Martin 
forgot that
that strategy of voting a full ranking even though one is indifferent
between some bottom candidates is a 0-info strategy. Public political
elections aren't 0-info elections.

If you expected that a certain candidate might be sincere CW, and
that order-reversal might be employed against him, then the
deterrent counterstrategy is to not rank anyone whom you like less
than him. The existence of this deterrent will deter offensive
order-reversal.

But even if it were a 0-info election (though it wouldn't be),
and a voter felt a temptation to increase his expectation slightly
by randomly ranking some bottom-end candidates, so what? Who is
wronged? That voter? He doesn't have to employ that strategy. A victim
of temptation?

Your claim that it will make him stay home because it's too much
work to randomly rank some equally-preferred bottom candidates
is more than a little silly. But evidently you're trying hard, and
you deserve credit for that.

--

Consequence Two: As noted, humans are *very* poor randomisers, even when 
trying hard to be random we are only 20% random. When not trying, we're 
about 2% random. The wrong candidate wins because of this, one who is 
sincere-rated last by all of the electorate.

Mike Comments: Again, you're forgetting that you're talking about a 
0-info strategy,
and that public political elections aren't 0-info. And again, who
is wronged because someone who is indifferent between some candidates
ranked them in an order that isn't random? Margins is much more
vulnerable to the voting of unfelt preferences than wv is.
For instance, show me how it violates majority rule or causes a
lesser-of-2-evils problem. Or any problem that someone has a right
to complain about. "No fair, my candidate lost because someone
ranked someone else over him when he really was indifferent between the
two!" [:-)] Oh you poor guy!

--

Consequence Three: Even if they are good randomisers, there's a random 
element. In small elections, that's a significant source of noise. The 
wrong candidate is elected as a result.

Mike Comments: Martin sets himself up as the arbiter of who is the right 
or wrong candidate to win.

But Margins doesn't elect the wrong candidate, does it, when it
dumps the sincere CW because mere truncation or easy order-reversal
has stolen the election [:-)]

--

Example of consequence two:

A,B,C,D, in that order on the ballot paper.

Sincere Votes and votes under margins:
10 B>A=C=D
10 C>A=B=D
10 D>A=B=C

Actual Votes under wv:
10 B>A>C>D OR B>A>D>C OR B>A(>C=D)
10 C>A>B>D OR etc
10 D>A>B>C OR etc

Result: A is the wv condorcet winner, despite being hated by everyone. 
Margins would give a draw to B,C,D.

Hence, winning-votes fails my newly invented Universally Despised 
Candidate Criterion, which makes it even worse than IRV's failure of 
Mike's Universally Unprefferred Candidate Criterion. Here A is hated, 
and hated by *everyone*, yet still gets elected.

Mike Comments: A is so much hated by everyone that they all consider him 
a 2nd choice and rank him alone in 2nd place [:-)]

In a real election, B,C, or D voters would estimate which of those
is the biggest rival, and try to him lose to the other 2 of {B,C,D}.
No, I haven't checked that out, because this example has just now
been posted, but at first glance it seems likely. That removes that
candidate from the tie, and improves your chance of your favorite
winning in the tie.

"Hated by _everyone_", who consider A a sincere 2nd choice, and
rank A alone in 2nd place  [:-)]

Serious enough for you?



More information about the Election-Methods mailing list